The Certified Corridor Feedback model code for Safe and Live LLM Guided UAV Navigation
Bildiriler > Tam Metin Bildiri
Yazarlar: Almamoori, Hayder
Zenodo (CERN European Organization for Nuclear Research), , , Uluslararası, Tam Metin Bildiri
Özet
Large Language Models (LLMs) provide a convenient interface for UAV mission planning by translating natural-language instructions into navigation goals. However, LLMs do not provide formal safety guarantees and may repeatedly generate unsafe or infeasible trajectories. This repository presents a closed-loop framework that combines: LLM-based waypoint planning High-Order Control Barrier Function (HOCBF) safety filtering Deadlock detection Control-to-language feedback Certified Corridor Feedback (CCF) The proposed approach guarantees collision avoidance through HOCBF-based safety filtering while restoring liveness through control-grounded feedback that guides LLM replanning. check README.md for more